1 files changed, 95 insertions, 19 deletions

diff --git a/fs/btrfs/send.c b/fs/btrfs/send.c
index 7ea2d6b1f170..dd38dfe174df 100644
--- a/fs/btrfs/send.c
+++ b/fs/btrfs/send.c
@@ -1160,7 +1160,6 @@ out:
 struct backref_ctx {
 	struct send_ctx *sctx;
 
-	struct btrfs_path *path;
 	/* number of total found references */
 	u64 found;
 
@@ -1213,8 +1212,6 @@ static int __iterate_backrefs(u64 ino, u64 offset, u64 root, void *ctx_)
 {
 	struct backref_ctx *bctx = ctx_;
 	struct clone_root *found;
-	int ret;
-	u64 i_size;
 
 	/* First check if the root is in the list of accepted clone sources */
 	found = bsearch((void *)(uintptr_t)root, bctx->sctx->clone_roots,
@@ -1231,19 +1228,6 @@ static int __iterate_backrefs(u64 ino, u64 offset, u64 root, void *ctx_)
 	}
 
 	/*
-	 * There are inodes that have extents that lie behind its i_size. Don't
-	 * accept clones from these extents.
-	 */
-	ret = __get_inode_info(found->root, bctx->path, ino, &i_size, NULL, NULL,
-			       NULL, NULL, NULL);
-	btrfs_release_path(bctx->path);
-	if (ret < 0)
-		return ret;
-
-	if (offset + bctx->data_offset + bctx->extent_len > i_size)
-		return 0;
-
-	/*
 	 * Make sure we don't consider clones from send_root that are
 	 * behind the current inode/offset.
 	 */
@@ -1319,8 +1303,6 @@ static int find_extent_clone(struct send_ctx *sctx,
 		goto out;
 	}
 
-	backref_ctx->path = tmp_path;
-
 	if (data_offset >= ino_size) {
 		/*
 		 * There may be extents that lie behind the file's size.
@@ -5082,6 +5064,7 @@ static int clone_range(struct send_ctx *sctx,
 	struct btrfs_path *path;
 	struct btrfs_key key;
 	int ret;
+	u64 clone_src_i_size;
 
 	/*
 	 * Prevent cloning from a zero offset with a length matching the sector
@@ -5107,6 +5090,16 @@ static int clone_range(struct send_ctx *sctx,
 		return -ENOMEM;
 
 	/*
+	 * There are inodes that have extents that lie behind its i_size. Don't
+	 * accept clones from these extents.
+	 */
+	ret = __get_inode_info(clone_root->root, path, clone_root->ino,
+			       &clone_src_i_size, NULL, NULL, NULL, NULL, NULL);
+	btrfs_release_path(path);
+	if (ret < 0)
+		goto out;
+
+	/*
 	 * We can't send a clone operation for the entire range if we find
 	 * extent items in the respective range in the source file that
 	 * refer to different extents or if we find holes.
@@ -5148,6 +5141,7 @@ static int clone_range(struct send_ctx *sctx,
 		u8 type;
 		u64 ext_len;
 		u64 clone_len;
+		u64 clone_data_offset;
 
 		if (slot >= btrfs_header_nritems(leaf)) {
 			ret = btrfs_next_leaf(clone_root->root, path);
@@ -5201,10 +5195,30 @@ static int clone_range(struct send_ctx *sctx,
 		if (key.offset >= clone_root->offset + len)
 			break;
 
+		if (key.offset >= clone_src_i_size)
+			break;
+
+		if (key.offset + ext_len > clone_src_i_size)
+			ext_len = clone_src_i_size - key.offset;
+
+		clone_data_offset = btrfs_file_extent_offset(leaf, ei);
+		if (btrfs_file_extent_disk_bytenr(leaf, ei) == disk_byte) {
+			clone_root->offset = key.offset;
+			if (clone_data_offset < data_offset &&
+				clone_data_offset + ext_len > data_offset) {
+				u64 extent_offset;
+
+				extent_offset = data_offset - clone_data_offset;
+				ext_len -= extent_offset;
+				clone_data_offset += extent_offset;
+				clone_root->offset += extent_offset;
+			}
+		}
+
 		clone_len = min_t(u64, ext_len, len);
 
 		if (btrfs_file_extent_disk_bytenr(leaf, ei) == disk_byte &&
-		    btrfs_file_extent_offset(leaf, ei) == data_offset)
+		    clone_data_offset == data_offset)
 			ret = send_clone(sctx, offset, clone_len, clone_root);
 		else
 			ret = send_extent_data(sctx, offset, clone_len);
@@ -6579,6 +6593,38 @@ commit_trans:
 	return btrfs_commit_transaction(trans);
 }
 
+/*
+ * Make sure any existing dellaloc is flushed for any root used by a send
+ * operation so that we do not miss any data and we do not race with writeback
+ * finishing and changing a tree while send is using the tree. This could
+ * happen if a subvolume is in RW mode, has delalloc, is turned to RO mode and
+ * a send operation then uses the subvolume.
+ * After flushing delalloc ensure_commit_roots_uptodate() must be called.
+ */
+static int flush_delalloc_roots(struct send_ctx *sctx)
+{
+	struct btrfs_root *root = sctx->parent_root;
+	int ret;
+	int i;
+
+	if (root) {
+		ret = btrfs_start_delalloc_snapshot(root);
+		if (ret)
+			return ret;
+		btrfs_wait_ordered_extents(root, U64_MAX, 0, U64_MAX);
+	}
+
+	for (i = 0; i < sctx->clone_roots_cnt; i++) {
+		root = sctx->clone_roots[i].root;
+		ret = btrfs_start_delalloc_snapshot(root);
+		if (ret)
+			return ret;
+		btrfs_wait_ordered_extents(root, U64_MAX, 0, U64_MAX);
+	}
+
+	return 0;
+}
+
 static void btrfs_root_dec_send_in_progress(struct btrfs_root* root)
 {
 	spin_lock(&root->root_item_lock);
@@ -6594,6 +6640,13 @@ static void btrfs_root_dec_send_in_progress(struct btrfs_root* root)
 	spin_unlock(&root->root_item_lock);
 }
 
+static void dedupe_in_progress_warn(const struct btrfs_root *root)
+{
+	btrfs_warn_rl(root->fs_info,
+"cannot use root %llu for send while deduplications on it are in progress (%d in progress)",
+		      root->root_key.objectid, root->dedupe_in_progress);
+}
+
 long btrfs_ioctl_send(struct file *mnt_file, struct btrfs_ioctl_send_args *arg)
 {
 	int ret = 0;
@@ -6617,6 +6670,11 @@ long btrfs_ioctl_send(struct file *mnt_file, struct btrfs_ioctl_send_args *arg)
 	 * making it RW. This also protects against deletion.
 	 */
 	spin_lock(&send_root->root_item_lock);
+	if (btrfs_root_readonly(send_root) && send_root->dedupe_in_progress) {
+		dedupe_in_progress_warn(send_root);
+		spin_unlock(&send_root->root_item_lock);
+		return -EAGAIN;
+	}
 	send_root->send_in_progress++;
 	spin_unlock(&send_root->root_item_lock);
 
@@ -6751,6 +6809,13 @@ long btrfs_ioctl_send(struct file *mnt_file, struct btrfs_ioctl_send_args *arg)
 				ret = -EPERM;
 				goto out;
 			}
+			if (clone_root->dedupe_in_progress) {
+				dedupe_in_progress_warn(clone_root);
+				spin_unlock(&clone_root->root_item_lock);
+				srcu_read_unlock(&fs_info->subvol_srcu, index);
+				ret = -EAGAIN;
+				goto out;
+			}
 			clone_root->send_in_progress++;
 			spin_unlock(&clone_root->root_item_lock);
 			srcu_read_unlock(&fs_info->subvol_srcu, index);
@@ -6785,6 +6850,13 @@ long btrfs_ioctl_send(struct file *mnt_file, struct btrfs_ioctl_send_args *arg)
 			ret = -EPERM;
 			goto out;
 		}
+		if (sctx->parent_root->dedupe_in_progress) {
+			dedupe_in_progress_warn(sctx->parent_root);
+			spin_unlock(&sctx->parent_root->root_item_lock);
+			srcu_read_unlock(&fs_info->subvol_srcu, index);
+			ret = -EAGAIN;
+			goto out;
+		}
 		spin_unlock(&sctx->parent_root->root_item_lock);
 
 		srcu_read_unlock(&fs_info->subvol_srcu, index);
@@ -6803,6 +6875,10 @@ long btrfs_ioctl_send(struct file *mnt_file, struct btrfs_ioctl_send_args *arg)
 			NULL);
 	sort_clone_roots = 1;
 
+	ret = flush_delalloc_roots(sctx);
+	if (ret)
+		goto out;
+
 	ret = ensure_commit_roots_uptodate(sctx);
 	if (ret)
 		goto out;